MetalloGen: Automated 3D Conformer Generation for Diverse Coordination Complexes

Conformer generation is crucial for computational chemistry tasks such as structure-based modeling and property prediction. Although reliable methods exist for organic molecules, coordination complexes remain challenging due to their diverse coordination geometries, ligand types, and stereochemistry. Current tools often lack the flexibility and reliability required for these systems. Here, we introduce MetalloGen, a novel algorithm designed for the automated generation of 3D conformers of mononuclear coordination complexes. MetalloGen accepts either SMILES strings or molecular graph representations as input and enables the generation of reliable conformers, including those with multiple polyhapto ligands, which are typically inaccessible to conventional conformer generators. To rigorously assess MetalloGen's performance, we benchmarked it on three distinct data sets: a curated collection of experimentally determined structures from the Cambridge Structural Database, the MOR41 benchmark set encompassing a wide range of organometallic reactions and complex ligand environments, and three catalytic reactions. Across all test sets, MetalloGen consistently reproduced appropriate geometries with high fidelity and demonstrated robust stereochemical control, even for challenging cases involving multiple polyhapto ligands. The versatility and reliability of MetalloGen make it a valuable tool for more accurate and efficient computational investigations in inorganic and organometallic chemistry.